Dan Hecht has posted comments on this change.

Change subject: Remove AMD Opteron Rev E workaround from atomicops
......................................................................


Patch Set 1:

> Just passing by. Re: "make out Atomic class use these functions" -
 > wouldn't we prefer to depend on std library atomic operations once
 > C++11 allows us to do so? Are these implementations superior?

I don't think we should use C++11 atomics directly, because they provide 
operator overloads and I think this makes code that relies on atomic operations 
too difficult to read (this is one problem with our current AtomicInt class 
which I plan to address).

One option will be to eventually base our new AtomicInt class on C++11, but I 
don't take it for granted that compiler writers get atomics correct/performant. 
 There's been a lot of issues with the various compiler support for atomics in 
the past.  The nice thing about gutil's atomic is it's easy to see what 
assembly code you get.  But, we can evaluate using C++11 atomics under the 
covers when we have C++11 support.  Let's discuss more in that patch (which I 
haven't yet posted).

-- 
To view, visit http://gerrit.cloudera.org:8080/2516
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I3639dcf86c14778967c0079b8dbc222a4516cf05
Gerrit-PatchSet: 1
Gerrit-Project: Impala
Gerrit-Branch: cdh5-trunk
Gerrit-Owner: Dan Hecht <[email protected]>
Gerrit-Reviewer: Dan Hecht <[email protected]>
Gerrit-Reviewer: Henry Robinson <[email protected]>
Gerrit-Reviewer: Michael Ho <[email protected]>
Gerrit-Reviewer: Tim Armstrong <[email protected]>
Gerrit-HasComments: No

Reply via email to